The Hardware Foundry
May 1, 2006 Timothy Prickett Morgan
Sometimes in life, you have to put up or shut up. You have to put your money where your mouth is. You have to lead, even if you don’t have the time. You have to try something that you think is right–or, might be right, but you’re not sure–even if a lot of other people say you are crazy, or if they listen quietly and are not enthusiastic about the idea because they don’t see what you do. This happens a few times in a good life, and that is what I face as I launch something I have called Hardware Foundry.
It is perhaps not an easy task to get a journalist to put his money where his mouth is, but I did that when I started the company that produces the IT Jungle stack of newsletters back in July 2001, after the company I was working for went quickly and completely bust. Back then, I started an IT newsletter publishing business because it was my best hope of doing the job I love to do: Write newsletters about the computer business. I also did it to employ some of the good people who were made unemployed by that company going bust. I made the decision to start a company almost on a whim, using gut instinct, only seconds after I learned I didn’t have a job. In all honesty, I do not make nearly as much money today as I did back in the IT boom-boom years–I often joke that I must be doing it for love, because I sure as hell am not doing it for money–and I am pretty sure that had I simply decided to get a journo job elsewhere, I would have made more money. But, as is often the case in jokes, it is true that I started my own company for love, not money. I love my employees, and I tell them that–often. They work hard, and I work hard right there in the trenches beside them, and together we have built a respectable publishing business, one that we plan to grow within the IT market and move beyond the IT market. The only kinds of plans I have are long term ones. I already have jobs mapped out for the grandchildren of all of us. (I’m kidding–sort of.) But IT Jungle was created not just for love and money, but also for control. All of us wanted and needed to have some control over our fates as professionals.
So it is with Hardware Foundry, which is also about control–specifically, the control of how pieces of hardware are designed and created, how the parts that go into making them are distributed and endorsed by users, and how we all make sense of myriad types of machines that get created. When you get right down to it, a manufactured object is nothing more than the sum of these parts. And I think that just as open source has been good for software–where the code, documentation, and specifications of code are put in the public domain–the same thing would be good for hardware design. What is good for an operating system or a database is equally as good for a PC, a server, a storage area network, a hammer, a car, a jet engine, a spaceship, or a wrench.
So many threads in my life have come together to convince me that the world needs something like Hardware Foundry. I have written a number of essays in the IT Jungle newsletters over the years, which I have posted on the Hardware Foundry site. This is certainly the nexus of the idea. I have complained about the lack of open benchmarks and a need for standard benchmarks. I have complained that political and economic forces that have nothing to do with the technical aspects of computers mean that IT vendors have more control over computer designs than the users who actually have to use the machines. I have complained in recent years specifically about the need for leaner and meaner designs that are focused on energy efficiency and that are not a slave to the rack form factors that were created two decades ago. I have said that someone out there should create a project that develops open source servers. Well, with Hardware Foundry, that is what I intend to do. And more.
But it all starts with computers. I was an early user of computers, even by modern standards. I am a lucky man in that I have several best friends, and one of them had an extremely smart and eccentric uncle (are these properties ever really separate?) who not only introduced me to my first real computer, but he gave it to me. It was an Ohio Scientific Z80 machine, which ran Basic and some CPM-like operating system; it was the size of two milk crates, and it had 8-inch floppy disk drives, which were all the rage. I never was much of a programmer, but I did a little coding on it. (You can read all about Uncle Dickie in the In Memoriam section of the Hardware Foundry site.) But perhaps more importantly, I was exposed to what a computer actually was, and what it could and could not do. I also learned what I found boring about computers, and what I liked about them. I like their guts. I like the way things plug together. I like making them work.
This love of making things work was the real reason I went to college to study engineering, and rather than let me indulge in that passion, Penn State, like other engineering schools the world over, bored me nearly to death with math and science when what I wanted to do was play with the toys and build things. Building things is the quickest way to understand how things work–at least for me. That is one reason why I quit engineering in my junior year and became an English major. They wouldn’t let me in the labs. But I never stopped wanting to be an engineer.
That’s why I was thrilled to be able to get my hands on some Hewlett-Packard ProLiant iron in the spring of 2003 and set up a data center for IT Jungle–well, it is actually a data closet more than a data center, as you can see from the picture. I knew the top brass at HP’s Industry Standard Server unit, and when I explained that I wanted to set up some servers in the closet of a New York City apartment with my own routers and T1 lines, at first they laughed. But I explained that the rack dimensions would fit in the closet, and further that my office is actually the kitchen of a studio apartment, and it has all the means of dumping the heat as well as 20-amp power lines (used for the electric stove). They laughed, and then they showed up with a half rack of servers, a storage area network, and some switches. The HP engineers were totally cool, and I half suspected that I was getting the iron from a failed dot-com. I didn’t ask any questions, and HP gave me about $60,000 in iron in trade for ads. That is the power of persuasion. I offered the same deal to IBM first, and then Dell, and both hemmed and hawed, and I got to say this to someone at both IBM and Dell:
“AHHHHHNNNN. Wrong answer. Next contestant, please.”
This is why the HP engineers were laughing, too. They knew that story, and they also knew the value of that “Powered by ProLiant” logo at the bottom of each IT Jungle page. They knew, as I did, that this was good marketing–especially since this iron was just sitting in a warehouse in New Jersey, collecting dot-com bomb dust.
Having that cluster of Linux and Windows servers has been quite an education for me. I love my ProLiant machines. But I also have needs, and they have not been met, and that is why HP has been monitoring its market share at IT Jungle pretty carefully. I have been on a kick to get low-power servers into the data closet, something that we did a few years ago. As I explained in the Lean Mean Green Machines essay I wrote for IT Jungle in February 2004, making smaller, cooler, and quieter servers was not just an idea to me. It was now something that was going to make my life demonstrably better. I nicknamed the IT Jungle cluster “Franklin,” because it was not only our printing press, but also a stove. And I kid you not, we dry wet clothes in the data closet, use it to help bread rise, and have used it to heat the apartment when it was bitter cold (we just turn off the exhaust to the outside world).
So I conned Justin Ward, my intrepid part-time IT manager, into helping me build a few of these Mini-ITX servers I was talking about being a possible idea. We built three of these boxes and put them into production, and they consume a lot less juice than a ProLiant. A lot less. And they are a little bit larger than a video cassette. They don’t have a lot of power, but if all you are doing is serving up flat HTML and a little PHP, then you don’t need much. That may sound primitive, but it is no accident. You have to make conscious choices about your software to drive down energy consumption on the hardware.
Ever since I wrote the Open Source Servers essay for IT Jungle in March 2005, my mind has pretty much been made up to do this. But little things helped convince me that I was right, that there should be a way for people to teach each other how to build things, and that it was important for us to do this.
For one thing, I couldn’t get the servers I wanted anywhere on the market. Something I complained about at length in the Open Source Servers essay. I was at LinuxWorld in Boston a year ago, doing an interview with John Fowler, the general manager of Sun Microsystems‘ Network Systems Group, the one that makes its X64 servers, and I was arguing with him that the then-future “Galaxy” machines should use the super-low-power Opteron EE and low-power Opteron HE chips as well as the standard parts so customers like me, with real tight power and heating budgets, could get the same performance we have, but burn a whole lot less juice. And he argued that the performance per watt doesn’t pan out that way, and I said I didn’t care, I wanted to burn less absolute watts and get the performance I needed, not maximize performance. I explained I had my own cluster in the house, running the IT Jungle site, and he said the most peculiar thing to me. “That’s kinda sad, actually,” he said to me, shaking his head.
Sad? To run a business? To actually be hands-on with the servers? Sad? This from a guy in charge of an X64 server division? I’ll tell you what sad is, John. Sad is you not getting my money because I am going to build the servers I want. I may not get it right the first or second time, but I will get it right. When Sun announced the “Niagara” Sparc T1 processors and related T2000 servers earlier this year, I told David Yen, the Sun executive in charge of its Scalable Systems Group, which makes Sparc-based servers, that me and my kids were taking on the whole Sun development team to deliver a set of servers with lower thermals than Sun’s designs. I bet him that we could do it. And, he laughed–but he didn’t take the bet.
But the issue I am trying to address with Hardware Foundry is larger than my own needs, and my desire to poke a stick at server vendors.
Here’s the larger problem. All of the people in the Western economies have been turned into consumers, myself included. I grew up poor enough that we had to rig and rejigger so many things that were breaking down all the time that you had to have a basic knowledge of how things work and a running inventory of junk stored up in the basement, garage, barn, and underneath the sink so you could quickly find or create a part or tool to fix whatever was going wrong. This, which may seem like hardship, is what makes life fun and challenging. But in a consumer economy, you don’t fix anything, and worse yet, most things are made so you can’t fix them and, even if you could, no one knows how anything works any more.
I’ll tell you a funny story. I am the president of a publishing company, and we put out eight newsletters a week. I also work as a consultant on my own and for Computerwire, a British publication that is owned by consultancy DataMonitor. My wife, Elizabeth, is a litigator with a big law firm here in New York, and we have two kids–Ellie, who is almost 7, and Henry, who is almost 5. We are busy people leading full and good lives. And to my great shame, for a short period of time, we hired a cleaning lady, who turned out to be an immigrant from Brazil. (I believe that if you are paying someone to clean your house or do your laundry, you either have too much money or you have lost control of your life. That’s the shame I am talking about.)
Anyway, Eleanie was wonderful. My house was spotless–at least on Tuesdays. She sang funny Brazilian songs as she worked. My sheepdog and my kids loved her. But she and the vacuum cleaner did not get along. One day, something went wrong with it, and I was busy trying to do my several jobs, working from the home office where the data closet is, and I walked out and said, “Damn. Now I have to buy another vacuum cleaner.” She shook her head at me and said, “You Americans. You just buy. In my country, we fix.”
At that point, I got angry enough to drop whatever I was doing, went to get my toolbox, and performed surgery on the vacuum cleaner. I took it completely apart, found the bent metal up the hose that was causing the blockage, and cleaned all the filters and every piece of the machine. I found things that were wrong that hadn’t caused problems yet, but would. And when I put it all back together and tested it, it didn’t work like it was new, but it was running a lot better than it had in a long time. I looked Eleanie in the eye and said, “In my country, we fix, too. So there.” To which she said one of the nicest things anyone ever said to me: “I wish my husband was just like you.”
Eventually, I let Eleanie go because, as I explained to my wife, I thought it was morally bad to have someone clean up after you, and that doing so teaches our children the wrong lesson. I really liked her, and we were all upset about it. But, I explained that it teaches the children that they can pay money to solve any problem, rather than picking up after themselves. So now, I give the kids an allowance, because I want to teach them about money and how to handle it; they get a bonus if they clean up their rooms after cartoons on Saturday, and they get a super-bonus if they do so without me reminding them. My mother was a bit more heavy-handed in her house when I was a kid. We all cleaned on Saturday, after cartoons and before lunch. No exceptions–even Dad. (Mom, that was a great rule, and I use it as the basis for a lot of rules I have in my life. And happy birthday today, too. My thanks to you and Dad for bringing me up right. I know it wasn’t easy dealing with me. I know because I have Henry–give me strength, God, give me strength. OK, I know revenge is sweet, but you don’t have to laugh that hard.)
About the same time I was wrestling with the vacuum cleaner last year, a kid in my co-op apartment building, named Marcel, who is one of the founding members of Hardware Foundry, bought a bunch of computer parts on eBay so he could build a computer. The parts were not under warranty, and suffice it to say, something was wrong in the parts and I did not have the tools or skills to figure out what. And that really bothered me. It was embarrassing that here I was, the president of an IT publishing company with a dozen computers in the house, and I could not figure it out. So I told Marcel that we were not only going to fix the computer, we were going to fix the larger problem.
Another thread: Craig Barrett and all of the top IT executives who constantly complain about how there are not enough indigenous engineers here in America, when I suspect what they really want is cheap, nearly indentured, technical labor. I single out Barrett because he gets under my skin. The top IT brass are always complaining about the lack of trained scientists and technicians, but they do not really do enough to bolster math, science, and engineering. (See IT Industry Execs Slam Education, Dodge Offshoring for why I get so angry at Barrett.) Both my wife and I are failed engineers, kids who loved math and science that the system failed (she studied physics at MIT for two years, then jumped to the University of Chicago to get a degree in medieval history). We wanted to be engineers, and we talk often about how we still do. But the system didn’t want us. It weeded us out. On my first day of engineering school, my first professor said this to me: “Look to your left, now look to your right. Those two people will not be here in three years.” Well, he sure was right. My aerospace engineering program weeded out 200 engineers out of 300. Just like the system wanted. And that is a crime. We should have been cultivated, not weeded out. It should have been more hands-on and less math. And the people that run engineering schools are morons for not seeing that. The head of the aero department at Penn State when I was there used to run NASA. ‘Nuff said. Draw your own conclusions.
Liz and I also talk a lot about our children, and how we want them to think about the world and their problems. We have been showing them math and science concepts since they could talk. My kids have been around computers–a lot of computers–from day one. And in my house, despite their multitudes, computers are a controlled substance. And the kids want their own machines. Badly. Just like all other kids do. So Ellie made her own, out of paper, when she was 4. I scanned it in so you could see it. She had the concept of the PDA. Soon after that, she made her own keyboard and monitor and system box and created a whole system, which she gladly sat down at and played with for about a week. But eventually, we got laptops, and the kids commandeer them from time to time to play educational games (I bought stacks of them) and to goof around playing other kinds of games whose educational merit is in question.
One day, in a fit of pique when my children were bothering me about getting their own computers and I had made my mind up to start Hardware Foundry because no other fool had taken my bait, I said this to them:
“You can have your own computers when you can buy the parts and build them. I will pay for the parts. But you will have to learn how to read first so you can figure out how to build them.”
To which Henry, who was maybe just 4 years old at the time said, “Show me.” As if it were a command, and I could just tap him three times on the forehead and he would know how to read. I explained that it would take a bit more effort for him to learn to read, but in the meantime I have some servers to build and I would love for them to both help. To which they shouted:
That, my friends, is how you make engineers. It starts here, right now. It starts by building, not just consuming. And with the Internet, you can design and build all over the world, with lots of people contributing all kinds of designs for all kinds of machinery. And you can cut the IBMs, the Dells, the HPs, and the Suns right out of the supply chain. We can do it ourselves, and give them a run for the–well, not for the money. A run for something else, then. Maybe a run for their lives.
I ask that you be patient with me as I launch the Hardware Foundry site. It is starting out modestly, with this essay and a bunch of others and a basic framework for what I want to do. I need to figure out the formats for hardware specifications, the tools people need to build things, the rules for running benchmarks on gear and providing pricing information, the ways that people and hardware parts suppliers can contribute, and a bunch of other stuff. This will take time.
All I know for sure is this: I get to be Chief Hardware Engineer.